Parallelizing bootscripts [was: Make bootscripts more POSIX compliant]

Dan Nicholson dbn.lists at gmail.com
Wed Feb 21 07:45:10 PST 2007


On 2/21/07, Ag. Hatzimanikas <a.hatzim at gmail.com> wrote:
>
> Parallelizing the bootscripts? -

I think that would be interesting and others have shown interest as
well. It would seem to be a pretty non-invasive change from the
current bootscripts.

> Using dash instead of bash? -

This was never really my goal. I personally feel like it's more
efficient to not use bash, but the bootscripts need to be POSIX
compliant. That needs to happen regardless. Or we change them to
#!/bin/bash.

But I don't want to change our default shell. Changing /bin/sh is an
option for the user.

> Using an alternative init system? Upstart?

That would be interesting, too, but obviously more invasive.

> I would like to see some more patches/commits on this,but the question is where?
> In the development branch?I don't think so.

DJ has his LSB style parallel bootscripts in the contrib directory in
bootscripts. So, people can start playing with those now and
contributing patches back. As it stabilizes, I guess we can chart a
course of action to make them the default. If it proves to be
reliable, of course.

The POSIX compliant stuff should go right into trunk. I submitted some
patches the other day that got things working on dash for me. I also
want to try posh, like Alexander suggested. There were a couple
changes that Bruce suggested, too.

If someone wants to investigate a big change like Upstart, then I
think that needs to go in its own branch, or someone will have to
maintain a private repo and submit back big patches.

> As it has been proposed in the past,a permanent experiment branch,it would help us a lot
> to speed and stabilize the development process,in nearly every aspect,before we start
> merge some serious changes such those.

Maybe. I'd rather have topic branches, though. Cause what would happen
if I had an upstart experiment going in this branch and then someone
came along and dumped a no-libc experiment in there? :-)

If there's significant interest in a big change, I'm all for creating
a branch for it. We probably wouldn't have got UTF-8 support in the
book without a branch. If the parallel boot scripting stuff gets
enough interest, then it will probably have to move to a branch for
full testing.

Proposals?

--
Dan



More information about the lfs-dev mailing list